Eliciting preferences for well-being measurement : methodological and behavioural challenges

Since the publication of the Stiglitz‑Sen‑Fitoussi Report in 2009, multidimensional and preference‑based measures of individual well‑being have received renewed attention. Measures grounded in the equivalence approach offer appealing normative properties, yet eliciting preferences for multidimensional well‑being remains challenging. Behavioural biases, statistical complications, and normative concerns have not been examined with sufficient rigour. This dissertation addresses several of these issues by testing existing methods, proposing new ones, and illustrating them in empirical applications. The overarching question is whether the equivalence approach can be operationalised in practice and whether it provides meaningful added value compared to less demanding well‑being measures. The contingent valuation method, a stated‑preference approach, can be used to elicit equivalent income—the income an individual would hypothetically retain after deducting her willingness to pay for perfect outcomes in all non‑monetary life dimensions. An online survey experiment using double‑bounded dichotomous choice questions shows that respondents’ willingness to pay is strongly and heterogeneously influenced by the first bid presented, indicating a substantial starting‑point bias. Moreover, analysis of the Belgian MEqIn data set reveals that equivalent incomes derived from contingent valuation do not outperform actual incomes in capturing multidimensional well‑being, likely due to behavioural biases. In contrast, the subjective well‑being method—estimating the impact of life dimensions on reported life satisfaction—is cognitively less demanding. A simulation study demonstrates, however, that estimating confidence intervals for the resulting well‑being measure requires appropriate statistical techniques. Among five methods compared, the widely used delta method performs the worst in terms of accuracy and robustness. Another statistical challenge concerns preference heterogeneity: imprecision leads to incomplete interpersonal rankings of equivalent incomes, and the degree of incompleteness increases with imprecision, heterogeneity, and the significance level. Additionally, the subjective well‑being method appears to underestimate the income coefficient, producing implausible equivalent income values and greater imprecision. Finally, the dissertation proposes the equivalent quantile measure as a normatively and empirically attractive alternative. It evaluates well‑being along the quantile path, which links outcomes across life dimensions quantile by quantile, forming a social ladder within society. Using the MEqIn data set, equivalent quantiles are shown to capture multidimensional well‑being more effectively than equivalent incomes, detect multidimensional deprivation, and remain robust to extreme preferences. Overall, the findings highlight that although preference‑based measures possess desirable normative features, they require highly precise and unbiased preference elicitation. Careful survey design and larger samples are essential, and several open questions remain before such measures can serve as reliable tools for policymaking.